Neuroticism in the digital age: A meta-analysis
Laura Marciano
*
, Anne-Linda Camerini , Peter J. Schulz
Institute of Communication and Health, Faculty of Communication Sciences, Universita Della Svizzera Italiana, Via Giuseppe Buffi 13, 6900, Lugano, Switzerland
A R T I C L E I N F O
Keywords: Internet activities
Problematic internet activities Personality
Neuroticism Meta-analysis Internet addiction Social media addiction
A B S T R A C T
The pervasiveness of the Internet has raised concern about its (problematic) use and the potentially negative impact on people's health. Neuroticism has been identified as one potential risk factor of Internet and other online addictions. To obtain a comprehensive quantitative synthesis of the association of neuroticism and both overall and problematic Internet activities, a meta-analysis was conducted according to the PRISMA guidelines. A comprehensive search was carried out in nine academic databases. After a stepwise screening procedure, 159 studies were eventually included: 104 studies in the meta-analysis on neuroticism and Internet activities and 64 studies in the meta-analysis on neuroticism and problematic Internet activities, comprising Internet, social media, Facebook, smartphone, and online gaming addiction (9 studies considered both aspects). When it comes to overall Internet activities, effect sizes were generally small and frequently non-significant, with some exceptions (e.g., expression of real me). For problematic Internet activities, a completely different picture emerged: high levels of neuroticism significantly correlated with all measures of problematic Internet activities, with medium size cor-relations. The differential results for Internet activities in general and problematic Internet activities can be related to problems in the conceptualization and assessment of the latter. More research is needed to overcome current conceptual and methodological issues and investigate the real nature of problematic Internet activities and to be eventually able to evaluate if neurotic people are really an at-risk population.
1. Introduction
Today, Internet pervasiveness has reached a plateau in different
affluent societies, with 90% of the U.S. (
Pew Research, 2019
) and 89% of
the European (
Eurostat, 2019
) population having access to it. At the same
time, access rates are growing in developing countries (
Pew Research,
2019
). Whether through a smartphone or a personal computer, Internet
access facilitates the exchange of information and interactions, for
example, through social media, and it offers sources of entertainment as
well as educational services. Despite beneficial outcomes of the Internet
such as increased social capital, peer and family connection, improved
self-expression and self-identity all associated with social media use
(
Bolton et al., 2013
), as well as increased civic engagement related to
news media consumption (
Pasek, Kenski, Romer,
& Jamieson, 2006
),
adverse outcomes have received far more attention in the scienti
fic and
public debate. In particular, the Internet and its problematic or excessive
use have been blamed for negatively affecting physical, mental, and
psychosocial functioning, especially in younger generations (
Ko, Yen,
Yen, Chen,
& Chen, 2012
).
Given the worrisome evidence on the detrimental effects of
problematic Internet use, scholars have invested their efforts in
identi-fying potential risk factors not only attributable to the Internet but also
the users. Neuroticism has been identified as a particularly significant
risk factor. Since neurotic people often experience negative feelings and
engage in maladaptive coping strategies during stressful situations
(
Carver
& Connor-Smith, 2010
), they can be more at risk of using the
Internet in an unhealthy way (
Hamburger
& Ben-Artzi, 2000
;
Hardie
&
Tee, 2007
). The considerable number of studies on neuroticism and
(problematic) Internet use calls for a comprehensive, quantitative
syn-thesis of the empirical
findings to provide evidence on the link between
the two. Hence, this study aimed to
fill the gap by summarizing the
existing literature on neuroticism and (problematic) online media
ac-tivities,
using
a
meta-analytic
approach.
More
precisely,
this
meta-analysis synthesized 159 studies investigating neuroticism, Internet
activities, and problematic or addictive Internet activities, As such, it is
an extension of previous syntheses of a much smaller number of studies
(e.g.,
Kayis¸ et al., 2016
). Results underline that neurotic persons,
although not engaging with the online world extensively, are more likely
to show signs of problematic online activities, including Internet, social
media, Facebook, smartphone, and online gaming addiction. This poses
* Corresponding author.
E-mail address:laura.marciano@usi.ch(L. Marciano).
Contents lists available at
ScienceDirect
Computers in Human Behavior Reports
journal homepage:
www.journals.elsevier.com/computers-in-human-behavior-reports
https://doi.org/10.1016/j.chbr.2020.100026
Received 5 February 2020; Received in revised form 25 June 2020; Accepted 4 July 2020 Available online xxxx
2451-9588/© 2020 Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
questions on whether problematic online activities actually trigger
addictive behaviors or behavioral problems which the person already
experienced in other (offline) settings. Methodological and conceptual
consideration of neuroticism and problematic online activities are further
addressed.
2. Literature review
2.1. Internet activities and Problematic Internet activities
While the concept of Internet use is rather well-de
fined and generally
includes the duration and frequency of the use of the Internet in general,
or specific contents and services, there has been a growing and still
un-solved debate about what constitutes problematic Internet use, also
called
“Internet addiction” (
Block, 2008
;
Shaw
& Black, 2008
). The lack
of consensus on the concept
– and the terminology used to describe it –
may be one of the reasons why Internet addiction has not (yet) been
inserted in the Diagnostic and Statistical Manual of Mental Disorders
(DSM-5,
American Psychiatric Association, 2013
) (
Block, 2008
;
Pies,
2009
), although Internet gaming disorder is now listed in the third
sec-tion of the DSM-5 and the diagnosis of Gaming Disorder has been
recently introduced in the 11th revision of the International
Classifica-tion of Diseases (ICD-11) (
World Health Organization, 2018
). However,
different terminologies such as
“addictive”, “pathological”,
“problem-atic”, “excessive” or “compulsive” Internet use reflect the unsolved
debate about the existence of a well-defined and widely acknowledged
Internet addiction or Internet disorder (
Shaw
& Black, 2008
). Thus, the
reader will
find the different terminologies used interchangeably during
the remainder of this meta-analysis to refer to the same concept as
described hereafter.
Both the DSM-5 and the ICD-11 describe
“addictive use” as a pattern
of repetitive behaviors, which can also be applied to characterize Internet
addiction and other, more specific forms such as social media addiction.
Internet addiction involves (i) a lot of time thinking about
Internet-related activities, even when not used (cognitive salience); (ii)
experi-ence of irritability and depressed mood related to not being online
(withdrawal); (iii) more time online than before to obtain the same
gratification(s) (tolerance); (iv) difficulties in regulating the time spent
on Internet activities (regulation problems), and (v) decreased time
dedicated to other hobbies and friends. Furthermore, symptoms are (vi)
persistent Internet use despite negative consequences at work, school, or
in the private sphere (persistent use) up to the point where (vii) the time
spent in online activities is hidden from others. In addition, symptoms of
Internet addiction include (viii) positive mood when being online, while
the mood changes to negative during offline periods (mood regulation),
which leads to (ix) a strong desire to engage in such activities (craving)
despite (x) interpersonal problems (e.g., with family members or friends,
at work or school) due to excessive use (
Black, Belsare,
& Schlosser,
1999
;
Jo et al., 2019
;
Shapira et al., 2003
;
Shaw
& Black, 2008
).
However, scholars have progressively questioned the concept of
Internet addiction, with critics targeting the lack of focus on specific
problematic online behaviors (
Starcevic
& Aboujaoude, 2016
), the
confusion between addiction to the Internet versus addiction on the
Internet (
Grif
fiths & Szabo, 2014
), the challenge in delimiting excessive
use as people are increasingly always online (
Starcevic
& Aboujaoude,
2016
), and the problematic separation between the Internet and the
device enabling access to it (e.g., smartphones, computer, tablets,
ap-plications). Further critics target the tricky differentiation between
proximal causes of Internet addiction, among others, anxiety and
depressive symptomatology, and the addiction itself (
Aboujaoude,
2010
). Additionally, in accordance to the criteria identified to detect
behavioral addictions, it seems that any online behavior, if carried out
excessively, can be described as problematic, or even pathological
(
Mihordin, 2012
). Indeed, the risk to categorized any exaggerated
attachment to the device as addictive is to neglect the pivotal
psycho-cognitive processes including motivations, cognitions, and
feelings that led and sustained the dysfunctional conduct (
Billieux,
Schimmenti, Khazaal, Maurage,
& Heeren, 2015
). Such shortcomings
lead to the lack of agreement upon proposed terminologies and
assess-ments of problematic Internet use or Internet addiction, and the absence
of a recognised de
finition of what constitutes Internet-related
psycho-pathology (
Aboujaoude, 2017
). Hence, although recent studies and
re-visions tried to disentangle different components of Internet (and other
media) addiction, a solution has not yet been found.
2.2. Personality and (problematic) internet activities
As stated by the Interaction of Person-Affect-Cognition-Execution
(I-PACE) model (
Brand et al., 2019
;
Brand Young, Laier, W€olfling, and
Potenza, 2016
), different factors contribute to the development and
maintenance of specific Internet-use disorders. In particular, a person's
characteristics (including bio-psychological factors, cognition,
personal-ity, existing psychopathology, and motives for media use) determine the
actual situation of a person, which can lead to using the Internet in a
certain way. The resulting Internet usage, in turn, elicits cognitive and
affective responses related to the grati
fications obtained, which foster its
subsequent use. The stabilisation of Internet use can diminish the
per-son's control of its use over time, leading to negative consequences in
daily life, which depend on the person's predispositions. Hence, one of
the crucial points to focus on when explaining (problematic) Internet use
are predisposing factors.
Among these factors, previous studies concentrated on a person's
personality, trying to explain if and how personality traits are related to
specific Internet use as well as problematic use. Using the Five-Factor
Model of personality traits as the main framework, previous reviews
and meta-analyses synthesized the relationship between neuroticism (or
emotional instability), extraversion, openness to experience,
agreeable-ness and conscientiousagreeable-ness (
Goldberg, 1990
), and (problematic) Internet
use (
de Francisco Carvalho, Sette,
& Ferrari, 2018
;
Kayis¸ et al., 2016
;
Liu
& Campbell, 2017
;
Twomey
& O'Reilly, 2017
), three reviews aimed to
infer personality from digital footprints (
Azucar, Marengo,
& Settanni,
2018
;
de Francisco Carvalho
& Pianowski, 2017
;
Tskhay
& Rule, 2014
),
and one focused on gaming disorder (
S¸alvarlı & Griffiths, 2019
). These
reviews and meta-analyses come to similar conclusions. For example,
based on twelve studies,
Kayis¸ et al. (2016)
reported correlational effect
sizes between Internet addiction and the Big Five personality traits and
showed that neuroticism was the only trait that was positively related to
Internet addiction. Another meta-analysis, based on four studies, focused
on problematic smartphone use and reported that it was associated with
neuroticism and impulsivity (
de Francisco Carvalho et al., 2018
). With
respect
to
general
(non-addictive)
media
activities,
only
one
meta-analysis (
Liu
& Campbell, 2017
) explored social media-related
ac-tivities and personality. Based on 33 studies, the authors found that
ex-traversion and openness were the strongest predictors of Social Network
Site (SNS) activities (e.g. SNS gaming, SNS interactions, posting photos,
number of friends, information seeking, and status update), whereas
agreeableness and conscientiousness showed few significant correlations
with social media use, and neuroticism correlated only with global social
media use. They concluded that
“researchers might want to focus at an
even narrower level than the Big Five traits by looking at aspects or
facets
” (
Liu
& Campbell, 2017
, p. 237).
2.3. Neuroticism and (problematic) internet activities
Given the results of past reviews and their recommendations for
future research, we decided to focus on neuroticism as a personality trait
that was not only found to be most frequently related to problematic
Internet (or media) use but also relevant from a public health perspective
(
Lahey, 2009
). Neurotic individuals tend to experience negative
emotional responses to challenges (
McCrae
& Costa, 2003
), and they are
often self-critical and sensitive to the criticisms of others, which foster the
perception of feeling inadequate (
Watson, Clark,
& Harkness, 1994
). In
general, core aspects of neurotic people involve negative affectivity (i.e.
the inclination to experience negative emotions), cognitive dispositions
(e.g., ruminative thoughts), abnormal reactivity to stress (e.g.,
psycho-physiological anxiety or distress), and behavioral or interpersonal
prob-lems (e.g., recklessness or hostility) (
Kirkegaard Thomsen, 2006
;
Suls
&
Martin, 2005
;
Widiger, 2017
). Growing evidence posits neuroticism as a
psychological trait of profound public health significance (
Friedman,
2019
;
Jeronimus, Kotov, Riese,
& Ormel, 2016
;
Lahey, 2009
;
Ormel
et al., 2013
). In fact, neuroticism correlates with and predicts many
distinct mental and physical disorders, including anxiety, mood, and
substance use disorders, as well as comorbidity among them and frequent
use of health services. High levels of neuroticism increase the risk of
having the most impairing and expensive mental health problems (
Lahey,
2009
). Neuroticism is also a reliable predictor of the quality of life and
longevity (
Friedman, 2019
), and a large amount of research (e.g.,
Jer-onimus, 2015
;
Ormel et al., 2013
;
Shackman et al., 2016
;
Van Os, Park,
&
Jones, 2001
;
Vink et al., 2009
) hints towards a theoretical explanation
called
“the vulnerability model”, stating that neuroticism promotes
processes that lead to common mental disorders (CMDs). More precisely,
high levels of neuroticism may lead to the development of CMDs both
directly and indirectly by increasing the negative impact of risk factors
such as stressful life events through cognitive processes such as the
negativity bias in attention, interpretation and recall of information, and
emotional processes such as increased reactivity and ineffective coping
(
Ormel et al., 2013
). Additionally, genetic overlaps have been found for
mood disorders and associated symptomatologies and neuroticism, i.e.
some genetic variants in
fluence both (
De Moor et al., 2015
).
Hence, understanding if and how neuroticism is associated with
Internet use and
– more importantly – problematic Internet activities or
addictions is today mandatory. However, a comprehensive quantitative
synthesis of the existing literature on neuroticism and (problematic or
addictive) Internet use is still missing. The present study aims to
under-stand better how neurotic people interact with the Internet, i.e., if they
tend to use it in a speci
fic way, and to which extent they are addicted to it.
For that purpose, we carried out a comprehensive meta-analysis,
including 159 studies on the relationship between neuroticism,
Internet activities, and problematic Internet activities. To our knowledge,
this up-to-date meta-analysis is the
first that aimed to comprehensively
understand the relation between these concepts and, in doing so, it
comprises far more studies than previous meta-analyses. In addition,
with respect to previous works, it distinguishes between Internet use
(overall and specific activities such as social networking) and
problem-atic Internet use, and it discusses differential relationships between these
two and neuroticism.
3. Methods
The meta-analysis was conducted according to the PRISMA guidelines
(
Shamseer et al., 2015
), which provide a practical guide on how
high-quality systematic reviews and meta-analyses should be carried out,
including instructions for each step (e.g., rationale and objectives, search
strategy, eligibility criteria, studies selection, data collection, risk of bias
assessment, and synthesis and discussion of results).
3.1. Search strategy and study selection
The research was carried out on February 7th, 2019 in titles of articles
listed in the following nine academic databases: Communication and
Mass Media Complete, Psychology and Behavioral Sciences Collection,
PsycINFO, PsychARTICLES, and CINAHL (all via EBSCOhost), ERIC and
ProQuest Sociology (both via ProQuest), Medline (via ProQuest, ISI Web
of Knowledge, and PubMed), and Web of Science. The search terms used
to identify eligible articles are reported in
Table 1
.
We imported all entries in Zotero to exclude any duplicates. Next, two
authors [L.M. and AL. C.] screened all titles and abstracts according to
prede
fined eligibility criteria and downloaded the full text of the retained
studies. We calculated Cohen's kappa statistic (
McHugh, 2012
) as a
measure of inter-coder reliability. Discrepancies after full-text screening
were resolved through a consensus meeting with the third author.
3.2. Inclusion and exclusion criteria
We included studies: (i) written in English, (ii) with a measure of
neuroticism (or emotional instability), (iii) published in a peer-reviewed
journal, (iv) with a measure of general Internet activities (including
on-line gaming) or a measure of problematic Internet use or onon-line
addic-tion, and (v) with a result convertible into an effect size. Studies reporting
information on consumer behavior, branding, shopping, learning,
multitasking, job contents, marketing, advertisement, cyberbullying,
validation, political contents, dating, sexting, therapies/counselling,
substance use, or digital footprint were excluded. Reviews, conference
papers, book chapters, and studies with an experimental or an
inter-vention design as well as studies that used the construct of anxiety as a
proxy of neuroticism were also excluded. We included thesis
disserta-tions when available online.
3.3. Data extraction and preparation
For each study that met the inclusion criteria, we collected
informa-tion on the following aspects: the country where the study was
con-ducted, study design (longitudinal or cross-sectional), sample size,
gender (% of male) and age of participants, type of recruitment (online vs
other), type of sampling procedure (random vs not), theoretical
back-ground (if reported), the personality scale used to measure neuroticism,
and the type of online (addictive) activity with its relative measurement.
Once we obtained a list with all the included studies, we grouped the
final results into two sections, summarizing conceptually comparable
Internet use behaviors into the same category: (i) general Internet use
(including online gaming) and (ii) addictive Internet use. A summary of
all the extracted information for each retained study can be found in
Appendix A
.
Table 1 Key terms used.
Context (in titles) Content (in titles) online OR chat* OR electronic OR mobile
OR social media OR social network* OR SNS OR SNSs OR OSN OR OSNs OR new media OR media use OR screen time OR digital OR instant OR messag* OR IM OR text-based media OR smartphone* OR online communication OR CMC OR ICT OR computer-mediated communication OR computer mediated communication OR smart phone* OR iPhone* OR facebook OR instagram OR snapchat OR selfie* OR posting OR internet OR TV OR television OR screen OR video* OR traditional media OR electronic media OR broadcast media OR internet addiction OR smartphone addiction OR social media addiction OR online game OR online gaming OR internet game OR internet gaming OR social network addiction OR problematic internet OR problematic smartphone OR problematic social media OR overuse internet OR excessive internet OR excessive smartphone OR excessive social media OR pathological internet OR pathological smartphone use OR pathological social media OR compulsive internet OR compulsive social media OR compulsive smartphone OR cellphone*
negative affect OR negative affectivity OR detachment OR introver* OR neurotic* OR negative emotional* OR emotional lability OR emotional instability OR emotional stability OR trait anxiety OR personality OR irritability OR bigfive OR five factor model OR FFM OR big5 OR big 5 OR IPIP OR TIPI OR irritability OR PID-5* OR PID-5-BF OR DSM-5 OR MMQ OR MPI OR EPI OR EPQ OR NEO-PI*
3.4. Methodological quality assessment
To assess the methodological quality of each study, we considered: (i)
the form of data collection (1
¼ online, 0 ¼ offline), (ii) the number of
items used to measure neuroticism, (iii) the presence of a reliability
check for multiple-item measures of neuroticism (1
¼ present, 0 ¼
ab-sent, NA
¼ not applicable due to single-item measures), and (iv) the
in-formation source for the Internet activity measure (i.e., 0
¼ self-reported,
1
¼ recorded from the web).
3.5. Meta-analytic procedure
The meta-analysis was carried out using the
“esc” (
Lüdecke, 2018
),
the
“meta” (
Schwarzer, Carpenter,
& Rücker, 2015
), and the
“metaphor”
(
Viechtbauer, 2010
) packages in R statistical software. We used Fisher's
r-to-z transformation as a measure of effect size, with results converted
back to r correlation coef
ficient for easier interpretation. We implied
different conversion formulas (
Bonett, 2007
;
Peterson
& Brown, 2005
;
Wan, Wang, Liu,
& Tong, 2014
) to convert raw data into the
final effect
size. We carried out several meta-analyses, one for each different
Internet-related activities. We interpreted the pooled effect according to
Cohen (1988)
, with r
¼ 0.10 as a small, r ¼ 0.30 as a medium, and
r
¼ 0.50 as a large effect size. To control for possible study diversities, we
used the inverse-variance method with a random-effects model and
Hartung-Knapp-Sidik-Jonkman adjustment (
IntHout, Ioannidis,
& Borm,
2014
) in each meta-analysis. The heterogeneity of the effect sizes was
calculated
with
the
between-study-variance
τ
2,
the
restricted
maximum-likelihood estimator (REML), and the I
2statistic (
Borenstein,
Hedges, Higgins,
& Rothstein, 2010
;
Higgins, Thompson, Deeks,
&
Alt-man, 2003
;
Ried, 2006
). According to the suggestions by
Higgins et al.
(2003)
, we interpreted the I
2level as low (25%), moderate (50%), and
high (75%). Potential publication biases were assessed via Egger's
regression test for funnel plot asymmetry (
Egger, Smith, Schneider,
&
Minder, 1997
). Finally, influence analyses (using the leaving-one-out
method) and meta-regressions (with bubble plots) were carried out to
investigate heterogeneity in the effect size. In meta-regression analyses,
we included as predictors the proportion (in %) of male participants and
the mean age. We employed meta-regressions only for meta-analyses
with a suf
ficient number of studies, which is preferably ten (
Higgins
&
Green, 2008
).
4. Results
The initial research produced 4238 entries, which resulted in 2095
after duplicates were removed. A
first title and abstract screening led to
the exclusion of publications on neuroticism and consumer behavior
(n
¼ 89), learning (n ¼ 40), digital footprint (n ¼ 30), political content
consumption (n
¼ 13), bullying and dating (n ¼ 9 and n ¼ 7 respectively)
as well as reviews (n
¼ 17) and conference papers (n ¼ 15). Additional
1533 publications were removed as they were out of topic (i.e., they did
not include a measure of neuroticism and/or (addictive) Internet use).
Full texts of the retained studies (n
¼ 386) were then screened by two
authors. At this stage, publications were excluded because they did not
include a measure of the constructs of interest or they did not provide raw
data to calculate the effect size (n
¼ 160), because full texts were not
available or they were books chapters, poster, or other duplicates
(n
¼ 62), or because they included clinical samples (n ¼ 5). The
remaining 159 articles were eventually included in the meta-analysis and
divided into two main categories reflecting conceptually comparable
Internet activities. In particular, 104 studies were included in the
meta-analysis on general Internet activities (including online gaming), and
64 studies were included in the meta-analysis on problematic Internet
activities and specific sub-categories (i.e., Internet, social media,
Face-book, smartphone, and online gaming addiction). Some studies were
included in both subcategories because they considered general and
addictive use. Cohen's kappa for the title and abstract screening for each
category was 0.708 for general Internet activities (including social media
activities), 0.739 for online gaming, and 0.836 for problematic Internet
activities, reflecting a substantial agreement between the coders
(
McHugh, 2012
). A comprehensive description of the screening process,
with reasons for publication exclusion, is reported in the PRISMA
flow-chart (
Fig. 1
).
4.1. Study description
Of the 159 included studies, only two studies, both conducted in
China and focusing on Internet addiction, applied a longitudinal design
(
Dong, Wang, Yang,
& Zhou, 2012
;
Zhou et al., 2018
), whereas all the
others (n
¼ 157) used a cross-sectional design. The majority of the studies
were published in peer-reviewed journals, and seven were dissertation
theses. Studies were conducted in North America, including the U.S.
(n
¼ 35) and Canada (n ¼ 5); Europe (n ¼ 46), including Germany
(n
¼ 10), United Kingdom (n ¼ 8), Poland (n ¼ 7), and Italy (n ¼ 6); Asia
and Oceania (n
¼ 41), including Australia (n ¼ 6), China (n ¼ 15), Korea
(n
¼ 9) and Taiwan (n ¼ 3); and the Middle East (n ¼ 25), including
Turkey (n
¼ 14) and Israel (n ¼ 6). One cross-sectional study investigated
inter-cultural differences by using samples from different countries,
including China, Greece, Israel, Italy, Poland, Romania, Turkey and the
U.S. (
Blachnio
& Przepiorka, 2016
). Sample sizes ranged from n
¼ 40
participants (
Amichai-Hamburger, Wainapel,
& Fox, 2002
) to n
¼ 21
0314
(
Gil de Zú~niga, Diehl, Huber, & Liu, 2017
), with a median of n
¼ 460.
Sixty-six studies recruited participants via online platforms, of which six
relied on single platforms such as Qualtrics or Amazon Mechanical Turk
(MTurk) to collect data. Only twenty-two studies used a random or
stratified sampling approach, and seventy-one implied a convenience
sample, mostly university students. The remaining 66 studies did not
specify the type of recruitment procedure.
Regarding the sample characteristics, the median percentage of male
participants was 44%, ranging from 0% (
Ezoe et al., 2009
;
Kakaraki,
Tselios,
& Katsanos, 2017
;
Molavi
& Pashaei, 2012
) to 97% (
Nithya
&
Julius, 2007
). The median age across all studies was 21.5 years (range
from 13 to 52 years). The majority of the studies included samples with
young adults (69% with an age range from 18 to 35 years). Fifteen studies
used an underage population (median
¼ 15.5). A summary of the
char-acteristics of all the included studies is reported in
Table 3
(
Appendix A
)
and
Figs. 2
–6
(
Appendix B
).
4.2. Measures of neuroticism
Neuroticism was labelled in different ways, however, the majority of
the studies (n
¼ 113) referred to the construct of “Neuroticism”, whereas
forty-one studies labelled it as
“Emotional Stability or Emotionality”, and
five studies used the updated construct of “Negative Affectivity” as part
of the new dimensional personality model proposed by the DSM-5
(
American Psychiatric Association, 2013
).
Table 5
(
Appendix B
)
pre-sents an overview of the measures utilised in the included studies,
including the number of items used to assess neuroticism. In general, the
following scales were applied the most: the Big-Five Inventory (BFI,
John
& Srivastava, 1999
; n
¼ 41 studies), including the 10-item version BFI-10
(
Rammstedt
& John, 2007
; n
¼ 18 studies), the Ten-Items Personality
Inventory (TIPI) (
Gosling, Rentfrow,
& Swann, 2003
; n
¼ 21 studies), the
Revised NEO Personality Inventory (
Costa
& McCrae, 2008
; n
¼ 19), the
Eysenck Personality Questionnaire (
Bodling
& Martin, 2011
; n
¼ 16),
and the International Personality Item Pool (IPIP,
Goldberg et al., 2006
;
n
¼ 16). The utilised scales included a different number of items to
measure neuroticism, ranging from two (of the BFI-10 and the TIPI) to 48
items (of the NEO PI-R and the Big-Five adjectives), however, studies
usually employed a ten-item scale, and 121 studies calculated a measure
of internal consistency.
4.3. Measures of internet activities
Internet activities were measured in different ways, however, the
majority of the studies focused on general Internet use (n
¼ 26). In
particular, they assessed the frequency of Internet use, the familiarity
with the Internet and the computer in general, as well as the use of the
Internet for information seeking, entertainment, and leisure activities.
The majority of the studies also reported information on online activities
through social media (n
¼ 53). To note, thirty-three studies were on
Facebook use, and only a minority was focused on other social media
platforms, such as Instagram, Pinterest, Twitter, Renren, and Yik Yak.
Usually, social media use was investigated not only in terms of having a
social media account, frequency of usage, and number of friends, but also
in terms of different activities. For example, some studies focused on
passive social media use, including information-seeking behaviors and
receiving feedback, and others on active social media use, including
self-disclosure (in terms of expression of real, hidden, and ideal me) and
social media use for communication. Some studies reported information
on taking selfies, and others investigated “faux pas” behaviors, i.e. the
use of social media to publish compromised material about the self.
Eleven studies focused on speci
fic activities of mobile phone use and app
usage. To note, one study (
Stachl et al., 2017
) collected information on
app usage directly from the Android phone of participants. Whereas the
majority of the included studies used self-reports to collect information
on media use, seven studies scraped the data directly from the social
media profiles of their participants, collecting information about the
number of friends, likes, posts, sel
fies about the participants’ profiles.
Concerning gaming behaviors, although a lot of the studies reported
results on the frequency of online gaming activities, some were also
focused on engagement in continuing to play, frequency of violent video
gaming, teamwork and socialisation on gaming platforms, and
engage-ment in creative tasks.
4.4. Measures of Problematic Internet activities
Problematic Internet use and equivalent concepts, such as Internet
addiction, social media addiction (including Facebook and Instagram
addiction), and,
finally, smartphone addiction, were assessed with 38
different scales. The Internet Addiction Test (
Young, 1998
) was the most
commonly used (n
¼ 23), and items from this scale were adapted in other
studies to measure Instagram and smartphone addiction or compulsive
use, problematic SNS use, and compulsive smartphone use. The Online
Cognition Scale by
Davis, Flett, and Besser (2002)
and the Chen Internet
Addiction Scale (
Chen, Weng, Su, Wu,
& Yang, 2003
) were utilised to
assess Internet addiction and problematic Internet use. Furthermore, the
Bergen Social Media Addiction Scale (
Andreassen, Torsheim, Brunborg,
& Pallesen, 2012
), including the dimensions of salience, tolerance, mood
modification, conflict, relapse, and withdrawal, was adopted to measure
social media and Facebook addictive behaviors. Facebook addiction was
also measured with the Problematic Facebook Use Scale, adapted from
the Generalized Problematic Internet Use Scale 2
– GPIUS-2 (
Caplan,
2010
), which addresses the core dimensions of the Cognitive-Behavioral
Theory (i.e. online interaction, mood regulation, cognitive
preoccupa-tion, compulsive use and adverse outcomes). Finally, the Smartphone
Addiction Scale (
Kwon et al., 2013
) was the most frequently scale utilised
to measure problematic smartphone use. Concerning online gaming, ten
studies focused on online gaming addiction, which was measured, among
others, with the Game Addiction Scale (
Lemmens, Valkenburg,
& Peter,
2009
) and the Assessment of Internet and Computer Game Addiction
(AICA-S- gaming) (
W€olfling, Müller, & Beutel, 2011
), an adapted gaming
version of the Generalized Problematic Internet Use Scale (
Caplan, 2002
)
(for an overview of all measures see
Appendix A
;
Table 4
).
4.5. Theoretical frameworks
Of the 159 included studies, only 9.4% (n
¼ 15) adopted a theoretical
framework. The Interaction of Person-Affect-Cognition-Execution
(I-PACE) model and the Uses and Grati
fication Theory were used more than
once (
Kircaburun, Alhabash, Tosuntaș, & Griffiths, 2018
;
Kircaburun
&
Griffiths, 2018
;
Laier, Wegmann,
& Brand, 2018
;
Wolniewicz, Tiamiyu,
Weeks,
& Elhai, 2018
). The I-PACE model is one of the most recent
at-tempts to explain specific Internet-use disorders (
Brand, Young, Laier,
W€olfling, & Potenza, 2016
). As described in the introduction of this
meta-analysis, the I-PACE model aims to give a comprehensive view of
the mechanisms underlying the development and maintenance of speci
fic
Internet-use disorders, with particular emphasis on the temporal
dy-namics of the addiction process. On the other side, the Uses and
Grati-fication Theory (UGT;
Katz, Blumler,
& Gurevitch, 1973
) focuses on
users’ motivations for and gratifications obtained from specific media
use. According to the theory, these two elements in
fluence the selection,
the frequency, and the intensity of media use (
Kircaburun, Alhabash,
Tosuntas¸,
& Griffiths, 2018
).
Table 6
(
Appendix B
) reports a list of all the
theories used.
4.6. Meta-analysis
All the meta-analytic results are summarized in
Table 2
, and Forest
plots are reported in
Appendix C.
4.6.1. Meta-analytic results for internet activities
Seventy-nine studies were included in the meta-analysis on Internet
activities. Overall, no significant results were found between neuroticism
and Internet/computer familiarity, Internet frequency, and online
in-formation seeking, whereas neuroticism signi
ficantly correlated with the
frequency of online leisure activities (n
¼ 8, r ¼ 0.040, 95% CI
[0.004-0.075], p
¼ .033, I
2¼ 51%, total sample ¼ 18,611), e.g. entertainment
contents, including music, videos, and images. With respect to social
media use, no significant relationship was found between neuroticism
and social media membership, number of friends, and feedbacks
received. However, neuroticism was positively related to the frequency
of social media use (n
¼ 25, r ¼ 0.090, 95% CI [0.049-0.131], p < .001,
I
2¼ 87%, total sample ¼ 28,000). Interestingly, neurotic people seem to
use social media more for self-disclosure than for communicative
pur-poses. In fact, we found a signi
ficant relation with social media disclosure
(n
¼ 22, r ¼ 0.071, 95% CI [0.016-0.125], p ¼ .013, I
2¼ 76%, total
sample
¼ 9130), expression of the real me (n ¼ 6, r ¼ 0.201, 95% CI
[0.024-0.365], p
¼ .033, I
2¼ 82%, total sample ¼ 1633), and a
margin-ally significant relation, probably due to the low number of studies, with
the expression of the ideal me (n
¼ 4, r ¼ 0.099, 95% CI [-0.01-0.205],
p
¼ .062, I
2¼ 53%, total sample ¼ 2190). We did not find any
relation-ships between neuroticism and the use of social media for
communica-tion (n
¼ 18, r ¼ 0.003, 95% CI [-0.037-0.043], p ¼ .870, I
2¼ 55%, total
sample
¼ 7965). Neurotic people seem to use social media in a more
passive way, as reflected by the significant relation with passive social
media use (n
¼ 11, r ¼ 0.055, 95% CI [0.009-0.101], p ¼ .024, I
2¼ 55%,
total sample
¼ 6426). Eventually, no significant relationships were found
between neuroticism and posting sel
fies, “faux pas” behaviors (i.e.
posting personal compromising material), and expression of the hidden
me. With regards to different online gaming activities, we did not
find
any signi
ficant relationships with neuroticism.
4.6.2. Meta-analytic results for Problematic Internet activities
Sixty-four studies provided effects sizes for the relationship between
neuroticism and problematic Internet use and other online addiction
measures (62 cross-sectional and 2 longitudinal). Meta-analytic results
showed that neuroticism signi
ficantly correlated with all online
addic-tion measures. In particular, neuroticism was found to be significantly
positively related to problematic Internet use, with a small-to-medium
effect size (n
¼ 34, r ¼ 0.249, 95% CI [0.201, 0.296], p < .001,
I
2¼ 92%, total sample ¼ 29,637). In a similar way, neuroticism
posi-tively and significantly correlates with social media addiction (n ¼ 12,
r
¼ 0.277, 95% CI [ 0.182, 0.368], p < .001, I
2¼ 88%, total
sam-ple
¼ 5949), Facebook addiction (n ¼ 8, r ¼ 0.217, 95% CI [0.161,
0.271], p
< .001, I
2¼ 75%, total sample ¼ 8779), and smartphone
Table 2 Meta-analytic results. Outcomes N k r CI [95%] Q I2 Internet activities General Computer familiarity 587 3 0.1165 [-0.224 .432] 5.42 63% Internet frequency 12,273 11 0.0570 [-0.0727; 0.1848] 322.97** 97%Online information seeking 21,266 10 -.009 [-.073; .056]
114.64** 92%
Online leisure activities 18,611 8 .040* [.004; .075] 14.42* 51% Online gaming Gaming frequency 28,675 10 .042 [-.095; .178] 142.98** 94%
Online gaming onset 21,935 3 .000 [-.225; .226]
38.10** 95%
Online gaming engagement 2798 3 .023 [-.184; .229]
6.51 69%
Online gaming teamwork 509 3 -.069 [-.436; .318]
7.39* 73%
Violent online gaming 763 3 .-.114 [-.504; .315]
16.11** 88%
Creative online gaming 417 2 .140 [-.734; .839]
2.95 66%
Social media activities
Social media membership 10,156 5 -.076 [-.341; .201]
581.18** 99%
Friends on social media 6998 10 .025 [-.032; .082]
50.52** 82%
Social media frequency 28,000 25 .090** [.049; .131] 187.71** 87% Expression of real me 1633 6 .201* [.024; .365] 27.75*** 82% Expression of hidden me 745 3 .219 [-.120; .513] 8.24* 76%
Expression of ideal me 2190 4 .099y [-.10; .205]
6.37y 53%
Receiving feedbacks 1684 3 .017 [-.321; .351]
17.50** 89%
Disclosure on social media 9130 22 .071* [.016; .125]
86.22** 76%
Social media passive use 6426 11 .055* [.009; .101] 22.42* 55% Communication on social media 7965 18 .003 [-.037; .043] 37.47** 55% Doing selfies 2031 4 .038 [-.121; .194] 12.33** 76%
Faux pas on social media 1275 3 .073 [-.189; .325]
5.16y 61%
Problematic Internet activities
Internet addiction 29,637 34 .249 ** [.201; .296]
410.39** 92%
Social media addiction 5949 12 .277** [.182; .368] 94.70** 88% Facebook addiction 9673 9 .154** [.005; .296] 246.28* 97% Smartphone addiction 5552 12 .296** [.224; .312] 30.16** 64%
Online gaming addiction 4935 10 .180** [.080; .276] 68.98** 87% Longitudinal correlations btw neuroticism at T1→ IA at T2 3384 2 .217y [-.055; .459] 1.38 28%
